Question 1

How much VRAM does LFM2.5-24B-A2B need?

Accepted Answer

LFM2.5-24B-A2B (24B) requires approximately 55.535527038574216 GB of VRAM at full precision (FP16), 27.8 GB at 8-bit quantization (Q8), and 13.9 GB at 4-bit quantization (Q4). The actual VRAM usage may be slightly higher due to KV cache and runtime overhead.

Question 2

What GPUs can run LFM2.5-24B-A2B?

Accepted Answer

We have 10 compatible GPUs indexed for LFM2.5-24B-A2B, including RTX Pro 6000 Blackwell, A100 80GB PCIe, A100 80GB, GH200, MI300X and 5 more. Compatibility is determined by whether the GPU's VRAM can fit the model's memory footprint, potentially using quantization to reduce memory usage.

Question 3

What is the cheapest GPU to run LFM2.5-24B-A2B?

Accepted Answer

The cheapest GPU to run LFM2.5-24B-A2B in the cloud is the RTX Pro 6000 Blackwell at $1.10/hr. Other affordable options include A100 80GB PCIe ($1.79/hr) and A100 80GB ($1.79/hr). Prices reflect the lowest on-demand rate across cloud providers.

Question 4

Can LFM2.5-24B-A2B run on consumer hardware?

Accepted Answer

Yes, LFM2.5-24B-A2B can run on consumer GPUs like the RTX Pro 6000 Blackwell (96 GB VRAM) at full precision (FP16).

Question 5

What quantization options are available for LFM2.5-24B-A2B?

Accepted Answer

LFM2.5-24B-A2B supports the following quantization levels: FP16 (full precision): ~55.535527038574216 GB, Q8 (8-bit): ~27.8 GB, Q6 (6-bit): ~20.8 GB, Q4 (4-bit): ~13.9 GB. Lower quantization reduces VRAM usage at the cost of some quality loss. Q8 offers a good balance between quality and memory savings, while Q4 provides maximum compression. As a Mixture of Experts (MoE) model with 64 experts (4 active), LFM2.5-24B-A2B has a larger total weight size but only activates a subset of parameters per token, which can affect effective memory usage during inference.

GPU	VRAM	Best Quant	From	Rent
RTX Pro 6000 Blackwell	96 GB	FP16	$1.10/hr
A100 80GB PCIe	80 GB	FP16	$1.79/hr	Rent
A100 80GB	80 GB	FP16	$1.79/hr	Rent
GH200	96 GB	FP16	$2.29/hr
MI300X	192 GB	FP16	$2.39/hr	Rent
H100 MEGA	80 GB	FP16	$2.89/hr	Rent
H100 SXM	80 GB	FP16	$2.89/hr	Rent
H100 NVL	94 GB	FP16	$2.89/hr	Rent
H100 PCIe	80 GB	FP16	$2.89/hr	Rent
MI350X	288 GB	FP16	$5.49/hr	Rent

LFM2.5-24B-A2B

VRAM Requirements

Compatible GPUs

LFM2.5-24B-A2B FAQ

VRAM requirements for LFM2.5-24B-A2B at different quantization levels
Quantization	VRAM Required
FP16	55.5 GB
Q8	27.8 GB
Q6	20.8 GB
Q4	13.9 GB